The extent of death clustering in rural Kenya: evidence from 1999 Kenya population and housing census

Abstract

This study set out to explore the extent of concentration of infant and child deaths in rural Kenya, as well as to investigate the factors that influence such clustering of deaths. It also hoped to come up with.' findings and recommendations th~would guide future programme planning and to influence policy on issues regarding child survival programmes. As a point of departure, the study was guided by two main research questions; is there evidence of death clustering in rural parts of Kenya? What are the factors that influence the concentration of such infant and child deaths? To be able to adequately address the above issues, I adopted a modified version of Mahadevan (1986) model of studying child survival. The adoption of this model and not the others popularly used in mortality studies e.g Mosley and Chen (1984), was premised upon two arguments. One is because the Mahadevan model unlike the Mosley and Chen one, has concepts and terminologies that are specifically related to mortality. This is important as it avoids the confusion that arises when using the term "intermediate" and "proximate" determinants, which are more related to fertility. Instead, Mahadevan has developed the terms "Life Affecting variables" and "imminent variables" to specifically refer to mortality determinants. Secondly, the Mahadevan model is flexible and enjoys a number of advantages that made it even more appropriate for this study. It incorporates both micro and macro variables whose influence on the health and consequent death of the child is diverse. Again the model clearly recognizes the fact that several LAV's (Life Affecting Variables) either in similar or dissimilar manner influence mortality in any society. It is this element of dissimilarity that made the model more appropriate for this study. Following the Mahadevan model (modified), two mam conceptual hypotheses were operationalized. These were that; environmental factors were likely to act via imminent variables to influence the risk of childhood deaths; and two, that child survival in Kenya is likely to be influenced by socio-cultural and socio-economic factors. During analysis to establish whether death clustering actually exists in rural Kenya, the study compared two maternal demographic characteristics i.e parity and age, as well as other characteristics i.e educational level attained and region of residence, with a theoretical statistical distribution i.e binomial. The purpose was mainly to assess the extent of differences in the variability among women and how these differences contributes to the observed child deaths after allowing for chance factors (Zaba and , David, 1996). Besides confirming the study hypotheses and answering the study objectives fully, the results also confirmed earlier findings of studies of a similar nature by other scholars (Khasakhala A. 1998; UNICEF 1998; Katende, J. 1983). A number of studies have shown that deaths, particularly infant and child deaths have a tendency to cluster or concentrate on certain women who share particular characteristics. These characteristics are but conditions which may range from Socioeconomic, socio-cultural, and environmental to demographic factors which determine the direction of magnitude such clustering may take (Das Gunpta, 1990; Khasakhala 1993; I McMurray 1997). This study has demonstrated that there is overwhelming evidence of death clustering among women in rural Kenya. This phenomenon is more pronounced among mothers of higher parity, and older ages (i.e 30 years and above). Secondly, the results have further indicated an over dispersion of the distribution of deaths between women in parity groups. The analyses have further shown that there is a true increase in risk concentration (and risk variability) among women at the highest parities within each age group - a real indication of death clustering.